Adversarial Actor-Critic Method for Task and Motion Planning Problems Using Planning Experience
We propose an actor-critic algorithm that uses past planning experience to improve the efficiency of solving robot task-and-motion planning (TAMP) problems. TAMP planners search for goal-achieving sequences of high-level operator instances specified by both discrete and continuous parameters. Our al...
| Main Authors: | , , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Published: |
Association for the Advancement of Artificial Intelligence (AAAI)
2021
|
| Online Access: | https://hdl.handle.net/1721.1/130053 |
| _version_ | 1826196412429238272 |
|---|---|
| author | Kim, Beomjoon Kaelbling, Leslie P Lozano-Pérez, Tomás |
| author2 | Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory |
| author_facet | Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory Kim, Beomjoon Kaelbling, Leslie P Lozano-Pérez, Tomás |
| author_sort | Kim, Beomjoon |
| collection | MIT |
| description | We propose an actor-critic algorithm that uses past planning experience to improve the efficiency of solving robot task-and-motion planning (TAMP) problems. TAMP planners search for goal-achieving sequences of high-level operator instances specified by both discrete and continuous parameters. Our algorithm learns a policy for selecting the continuous parameters during search, using a small training set generated from the search trees of previously solved instances. We also introduce a novel fixed-length vector representation for world states with varying numbers of objects with different shapes, based on a set of key robot configurations. We demonstrate experimentally that our method learns more efficiently from less data than standard reinforcementlearning approaches and that using a learned policy to guide a planner results in the improvement of planning efficiency. |
| first_indexed | 2024-09-23T10:26:27Z |
| format | Article |
| id | mit-1721.1/130053 |
| institution | Massachusetts Institute of Technology |
| last_indexed | 2024-09-23T10:26:27Z |
| publishDate | 2021 |
| publisher | Association for the Advancement of Artificial Intelligence (AAAI) |
| record_format | dspace |
| spelling | mit-1721.1/1300532021-09-10T19:55:28Z Adversarial Actor-Critic Method for Task and Motion Planning Problems Using Planning Experience Kim, Beomjoon Kaelbling, Leslie P Lozano-Pérez, Tomás Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science We propose an actor-critic algorithm that uses past planning experience to improve the efficiency of solving robot task-and-motion planning (TAMP) problems. TAMP planners search for goal-achieving sequences of high-level operator instances specified by both discrete and continuous parameters. Our algorithm learns a policy for selecting the continuous parameters during search, using a small training set generated from the search trees of previously solved instances. We also introduce a novel fixed-length vector representation for world states with varying numbers of objects with different shapes, based on a set of key robot configurations. We demonstrate experimentally that our method learns more efficiently from less data than standard reinforcementlearning approaches and that using a learned policy to guide a planner results in the improvement of planning efficiency. NSF (Grants 1523767 and 1723381) AFOSR (Grant FA9550-17-1-0165) ONR (Grant N00014-18-1-2847) 2021-03-02T19:31:01Z 2021-03-02T19:31:01Z 2019-07 Article http://purl.org/eprint/type/ConferencePaper 2374-3468 2159-5399 https://hdl.handle.net/1721.1/130053 Kim, Beomjoon et al. "Adversarial Actor-Critic Method for Task and Motion Planning Problems Using Planning Experience." Proceedings of the AAAI Conference on Artificial Intelligence 33, 1 (July 2019): 8017-8024 © 2019 Association for the Advancement of Artificial Intelligence http://dx.doi.org/10.1609/aaai.v33i01.33018017 Proceedings of the AAAI Conference on Artificial Intelligence Creative Commons Attribution-Noncommercial-Share Alike http://creativecommons.org/licenses/by-nc-sa/4.0/ application/pdf Association for the Advancement of Artificial Intelligence (AAAI) MIT web domain |
| spellingShingle | Kim, Beomjoon Kaelbling, Leslie P Lozano-Pérez, Tomás Adversarial Actor-Critic Method for Task and Motion Planning Problems Using Planning Experience |
| title | Adversarial Actor-Critic Method for Task and Motion Planning Problems Using Planning Experience |
| title_full | Adversarial Actor-Critic Method for Task and Motion Planning Problems Using Planning Experience |
| title_fullStr | Adversarial Actor-Critic Method for Task and Motion Planning Problems Using Planning Experience |
| title_full_unstemmed | Adversarial Actor-Critic Method for Task and Motion Planning Problems Using Planning Experience |
| title_short | Adversarial Actor-Critic Method for Task and Motion Planning Problems Using Planning Experience |
| title_sort | adversarial actor critic method for task and motion planning problems using planning experience |
| url | https://hdl.handle.net/1721.1/130053 |
| work_keys_str_mv | AT kimbeomjoon adversarialactorcriticmethodfortaskandmotionplanningproblemsusingplanningexperience AT kaelblinglesliep adversarialactorcriticmethodfortaskandmotionplanningproblemsusingplanningexperience AT lozanopereztomas adversarialactorcriticmethodfortaskandmotionplanningproblemsusingplanningexperience |